Regarding capacitors:
I haven't really measured and compared different configurations to determine how many capacitors are actually needed, but in general it is better to have more (the downside is that you get a larger spark, but that's what anti-spark switches are for). Having more capacitors will also put less wear on them since the ripple is shared among them and the total ESR gets lower. If you really need to save some space/cents you can try using less capacitance, but I will continue using about the same value.
FOC does not actually output sine waves, it is still switching but it modulates sine waves which together with the motor inductance makes the current more sine wave shaped. How much capacitance is needed to deal with the ripple still depends on how fast switching is done, so high frequency trapezoidal commutation might require less capacitance than low frequency FOC if the difference in frequency is large, even though the modulated current has less ripple with foc. However, in general, FOC will do more switching and require less capacitance. More switching also means more switching losses, which means the the FETs will get warmer when running foc. I have already confirmed that this is the case in my initial tests, which was expected.
Dr_T said:
Thanks for posting that link to SLS Lizard! finally had some time to browse it through. Nothing beats some datas! From what I am trying to understand from FOC so far (which might not be a whole lot yet ), the higher efficiency at partial load, lower ripple / lower stress on batteries and caps is all very welcome for our fast RC car toys
.
@Vedder: you mentioned having more trouble driving a low inductance inrunner (Lizard's 1717?) on FOC than the typical sub 200kv ~12 pole ~60mm outrunners. Is there a way to recognize motors that will work well on FOC, based on their basic properties?
Also, I recently got a ironless 44x85mm 2-pole ~1000kv inrunner I am wanting to try with the VESC; does being ironless make it easier or worse to drive for VESC, or does it not matter at all?
Thanks!
Coreless motors have less inductance in general and are more difficult to use. Also, the coreless motors I have tested tend to be inefficient in general, but maybe that is because of low quality. However, 1000kv and two poles means that the motor has low electrical speed, which is a good thing. You should be able to run that motor at full 12s.
Regarding RC motors and FOC, here a blog post from TI:
https://e2e.ti.com/blogs_/b/motordrivecontrol/archive/2013/11/07/you-will-not-be-able-to-spin-this-motor
They write how their customer has an extremely difficult motor with 9.2µH inductance and that they made it run. Yesterday I measured my 4-pole 2100kv RC inrunner - it has 2µH, which is less than one quarter of that
It did work with the FOC implementation though, but the estimated current was noisy and off because of the low inductance. I'm surprised that it works. I still need to test this more and under load though, so I will try it on lizards motor some time in the next days.
zener said:
Dear Vedder,
Please considder to implement (after the FOC integration) the new very awesome IPT007N06NATMA1 MOSFET in your next version of VESC. It costs only 1.41 more and is available from 9.12.
Here is comparison with the current MOSFETS IRFS7530TRL7PP
Thanks, zener
Thanks, I will have a look at that. 1mOhm typical and 0.75mOhm is not a huge difference though.
I was hoping to release a beta version of the FOC implementation a bit sooner, but I decided to keep it a bit longer before I release it. The reason is that I want it to work well enough to be useful, and I want to have some auto-detection of motor parameters so that most people can use it without a whole lab to measure their motor parameters. For the sensorless observer it is very important to have correct values of the motor resistance, inductance and flux linkage, and the only parameter that hobby motor manufacturers might specify is the resistance, and it usually is wrong.
As of yesterday I have an automatic estimator of resistance and inductance that works for most of my motors, and the flux linkage estimator is about the same difficulty to use as the old motor detection implementation. I will add that to BLDC tool in the next days, test everything a bit more and hopefully push it to github within a week (I said that two weeks ago as well).
Don't expect any magic from FOC. You will hear a big difference in sound, but you are not likely to gain any noticeable difference in efficiency. The FETs will most likely get warmer and the motor efficiency might increase a bit a low speed, which can be good for hub motors and ebikes. Also, the top speed (or motor kv) will get lower with FOC with a factor of sqrt(3) / 2 or 0.866 since the modulation is a bit limited when full sine waves are modulated. With a three shunt version of the VESC the modulation could be increased to 100%, but that means that the commutation at full speed will be trapezoidal, the same as bldc. Another way to increase the top speed is using field weakening, but I don't know how well that works and how efficient that is on our motors. The difference in sound is really worth FOC for me though